Designer’s Guide Community

Bhupendra

Junior Member

Offline

Posts: 13

effective length less than or equal to zero?
May 03^rd, 2009, 12:55pm

Hi all,

I recently installed the AMI 0.6u NCSU design kit (C5N) and did a small simulation.

While spectre is able to generate the netlist it crashed and throws the following error:

Error found by spectre during initial setup.
�ERROR (CMI-2213): I0.M0: Effective length is less than or equal to zero.

I checked the device length and it is set to 900n (min L = 600n)

Has anybody come across a similar problem while working the NCSU CDK?

I am in a fix right now as I cannot proceed from here on.
Any help/pointers in this regard would be really appreciated.

Regards,
Bhupendra

Ps.
The output log file is attached for reference.

Back to top

IP Logged

Bhupendra Junior Member Offline Posts: 13	Re: effective length less than or equal to zero? Reply #1 - May 3^rd, 2009, 12:57pm Just realized the forum would not allow me to attached txt files. Attached is a snapshot of the log file. -Bhupendra
Back to top	Picture_1_001.png IP Logged

Geoffrey_Coram

Senior Fellow

Offline

Posts: 1999
Massachusetts, USA

Re: effective length less than or equal to zero?
Reply #2 - May 4^th, 2009, 7:49am

Check "input.scs" to make sure the 900n actually got there -- and that it didn't get scaled (eg, maybe the schematic entry tool expects values in microns, so you have (900n) um, which would be 900e-15m, and then a delta-L term could make it zero.

Try a longer drawn length, maybe delta-L is bad?

Try a different width.

Back to top

If at first you do succeed, STOP, raise your standards, and stop wasting your time.

IP Logged

Bhupendra

Junior Member

Offline

Posts: 13

Re: effective length less than or equal to zero?
Reply #3 - May 4^th, 2009, 8:01am

Hey Geoffrey,

Thanks a lot for the response.
You nailed the exact problem here!
That is precisely the reason. I figured it out yday itself.
Sorry for updating it so late in here for everyone else.

Turns out the schematic editor translates the W and L in micron so if I enter 1u it actually translates it as 1e-12 and hence the error.

Attached is the property window of one of the MOS. Here we enter the dimensions as W=3 and L=1 and the editor translates it to 3u and 1u respectively.

The interesting thing however was that irrespective of how we enter the W and L the input.scs (netlist) shows the dimensions correctly!

That brings me to couple of questions:

1) the W/L paramters (as in the picture) have # grids units associated with it. And in this current scheme it becomes humongous! Does that mean anything in terms of layout/pcells?

2) Is there a way I can set things the old fashioned way. Just enter the units along with the number?

-Bhupendra

Back to top

Picture_2_001.png

IP Logged